Age Encoded Adversarial Learning for Pediatric CT Segmentation
Related Experiment Video
Updated: Jun 27, 2025
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
Saba Heidari Gheshlaghi1, Chi Nok Enoch Kan2, Taly Gilat Schmidt3
1Department of Computer Science, Marquette University, Milwaukee, WI 53233, USA.
Insights
This study introduces a new deep learning framework, CFG-SegNet, to improve organ segmentation in pediatric CT scans, addressing data limitations. The method enhances accuracy for organs like the liver and heart, crucial for medical diagnoses.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer Vision
Background:
- Accurate organ segmentation in CT images is vital for disease diagnosis, treatment planning, and radiation therapy.
- Data scarcity, especially in pediatric CT segmentation due to radiation sensitivity, poses a significant challenge.
- Existing methods struggle with limited datasets, impacting diagnostic and therapeutic precision.
Purpose of the Study:
- To develop a novel segmentation framework, CFG-SegNet, to overcome data limitations in pediatric CT organ segmentation.
- To incorporate an auxiliary classifier generative adversarial network (ACGAN) that conditions on age for enhanced feature generation.
- To improve the accuracy and robustness of organ segmentation in low-data scenarios.
Main Methods:
- Proposed a conditional feature generation segmentation network (CFG-SegNet) integrating an age-conditioned ACGAN.
- Utilized a single loss function and 2.5D segmentation batches for training.
- Experimented on a dataset of 359 pediatric subjects (5 days to 16 years).
Main Results:
- CFG-SegNet achieved high average Dice Similarity Coefficients (DSC): 0.681 (prostate), 0.619 (uterus), 0.912 (liver), and 0.832 (heart).
- Demonstrated improved segmentation accuracy compared to U-Net by 2.7% (prostate), 2.6% (uterus), 2.8% (liver), and 3.4% (heart).
- The framework showed superior performance in segmenting organs with limited training data.
Conclusions:
- The proposed CFG-SegNet effectively addresses data limitations in pediatric CT organ segmentation.
- The age-conditioned ACGAN enhances feature generation, leading to more precise segmentation.
- This framework offers a promising solution for improving medical image analysis in data-scarce pediatric populations.
Abstract:
Organ segmentation from CT images is critical in the early diagnosis of diseases, progress monitoring, pre-operative planning, radiation therapy planning, and CT dose estimation. However, data limitation remains one of the main challenges in medical image segmentation tasks. This challenge is particularly huge in pediatric CT segmentation due to children's heightened sensitivity to radiation. In order to address this issue, we propose a novel segmentation framework with a built-in auxiliary classifier generative adversarial network (ACGAN) that conditions age, simultaneously generating additional features during training. The proposed conditional feature generation segmentation network (CFG-SegNet) was trained on a single loss function and used 2.5D segmentation batches. Our experiment was performed on a dataset with 359 subjects (180 male and 179 female) aged from 5 days to 16 years and a mean age of 7 years. CFG-SegNet achieved an average segmentation accuracy of 0.681 dice similarity coefficient (DSC) on the prostate, 0.619 DSC on the uterus, 0.912 DSC on the liver, and 0.832 DSC on the heart with four-fold cross-validation. We compared the segmentation accuracy of our proposed method with previously published U-Net results, and our network improved the segmentation accuracy by 2.7%, 2.6%, 2.8%, and 3.4% for the prostate, uterus, liver, and heart, respectively. The results indicate that our high-performing segmentation framework can more precisely segment organs when limited training images are available.
